Longitudinal data from two cohorts of Roman Catholic sisters (nuns) will be used to test hypotheses relating demographic-, cognitive-, and family mortality-related measures during early and mid life to longevity and to independence and dependence in activities of daily living (ADLs). The Historic Cohort consists of 1,975 sisters born between 1890 and 1915 who joined a religious congregation at about age 20. The Surviving Cohort consist of an estimated 760 sisters from the "Historic Cohort" who survive to 1990 and agree to participate in four annual assessments of mental and physical function and ADLs during 1990-1994. Archival information and previously collected questionnaire data will be used to characterize early and mid life exposures. Life tables will estimate life expectancies and the mean duration of independence and dependence in each ADL for various "exposure" groups. Logistic regression will estimate for individual independent variables the multivariate-adjusted odds of sisters "surviving to old age (1990) with intact ADLs" as well as the odds of 75 to 100 year- old sisters developing "ADL-dependencies" during 1990-1994. Latent variable structural equation analysis (LISREL) will examine hypothesized patterns of relationships of the early and mid life measures with one another and with various health outcomes. Hypotheses to be tested: (1) Each year of attained education will be associated with approximately one extra year of life expectancy after age 20, and a significant increase in the odds of 20-year-old sisters "surviving to old age with intact ADLs". Each year of education is associated with a moderate increase in the duration of "ADL-independence", a small decrease in the duration of "ADL- dependence", and a moderate decrease in the odds of developing "ADL- dependencies". (3) Relationships between education and the health outcomes described above will be explained, in part, by five correlates of education (which each make independent contributions to the health outcomes): socioeconomic status of the family of origin, mortality of the family during the subject's childhood, writing skills at about age 20, occupation, and frequency of holding elected offices in the congregation. Because of the relatively homogenous adult lifestyle of the study population, our findings should not be confounded by factors such as smoking and alcohol use, reproduction history, marital status, living arrangements, income, social isolation, nor materially affected by changes in the prevalence of these characteristics across age groups or birth cohorts. This is a well defined, cooperative, and essentially closed population in which to analyze survival and the loss of ADLs.